Data science applies advanced techniques from mathematics, statistics, computer science, and related fields to analyze large data sets. In business, a data science problem is almost always an optimization problem, not just an analysis problem. That is, the task is not merely to discover interesting patterns, but rather to find the best possible way to do something.
An organization has big data when it has too much data to store and process economically in traditional data stores, notably relational databases. Often big data arises from business activities on the Internet. Businesses use noSQL and newSQL technologies to manage big data, instead of relational databases, often using data science to analyze their big data. Data science is not limited to big data. Scientists have been analyzing small data sets for a long time!
Business intelligence is a mature collection of practices centered on building relational databases that contain historical data. There are several kinds of historical databases: data warehouses, data marts, operational data stores, and data cubes. One can use data science to analyze data stored in any of these database architectures. Data science begins where traditional analytical techniques based on querying (such as online analytical processing) end.
That’s an interesting question, and we devote an entire blog post to addressing it. Here’s the short version.
Traditional experts’ domain knowledge is far less specific than a good data science model. Also, most of the time traditional expertise is the result (at best) of dated analytical methodologies that are less powerful than those a data scientist can bring to bear. As a result, a data scientist should reach beyond traditional domain knowledge and take a fresh look at the business problem, casting it in the most specific terms possible (how will a single consumer, stock, motor vehicle, airplane, assembly line, etc. behave? how do we most profitably manage that one specific case?), and applying the most powerful computational methods possible. The result will usually be a significantly more powerful and beneficial model than an industry expert possesses. A data scientist should be expert about data science, not about any particular industry. Mosaic’s experts are as deep as they come. They have to be, to do data science for NASA in the air-traffic management (ATM) industry. ATM problems are very hard, and NASA is a very demanding customer.
Mosaic has a decade of experience solving complex data science problems for NASA, the FAA, several airlines, and the shipping industry. Our practice has focused on these customers’ data science challenges because they are among the most challenging problems in commercial data science. About 50,000 commercial flights cross the United States every day, taking off and landing at over 500 commercial airports. Every few seconds radar installations around the country track each flight in the air, generating large volumes of data every day. These large data sets combine with variable supply and demand patterns, weather uncertainties, very high operational costs, interactions with civilian (general aviation) and military flights, and risks involving many human lives, to create many opportunities for making air travel safer, faster, and less costly. As a result, our projects routinely demonstrate or deliver savings of several million dollars per year, as well as substantial delay reductions, safety improvements, and increased throughput. Our projects also require that we build highly accurate air-traffic simulations, which we use to develop and test our data science models, to demonstrate thoroughly that the models we build actually work, long before our customers deploy those models in the real world. All of which makes for very interesting data science!
Mosaic is small, but it’s very stable, and has served large customers for a decade. Some of our data scientists have three decades of industry experience.
While we’re small, our expertise is solving large, data-intensive business problems. Mosaic approaches each data problem with an open mind, keen to understand and make the most of a business opportunity on its own terms. That flexibility lets us focus on the optimization problem. (Technology is the easy part.) When you see an opportunity for your company to achieve large-scale improvements by using data systematically, Mosaic has the depth of experience and expertise you need!
Some young data science consultancies have rosters of PhDs in a single academic discipline (physics, say), or a narrow range of such disciplines. Mosaic’s experience solving optimization problems for large, established logistics operations has led us to take a very different tack. Beyond our strong focus on optimization, we bring a diverse mixture of disciplines to each problem, including cognitive science, human factors, data architecture, organizational behavior, and process engineering. We even have a full-time meteorologist on staff, to guide our modeling of weather uncertainties. This diversity helps us account for every important aspect of a business problem, so that the solution we deliver works in the real world. We’ll make your data soar, but we’ve got both feet on the ground.
Curious how well that approach works? Read our case studies.
This question has a long answer, which we’ve documented in our white paper “Standing up a Data Science Group.” The short answer is, perhaps you should do it yourselves—at least part of it. Enterprise-scale data science has a long learning curve for both individuals and organizations. If your ambition is eventually to do it yourselves, we can substantially shorten your learning curve, accelerate your delivery on early projects, and reduce early risks of failure, by providing coaching and leadership on your early efforts. Experience is a dear school!
Mosaic’s solutions routinely demonstrate or deliver multi-million dollar annual savings. We don’t just crunch numbers. Rather, we take a very scientific approach to working with your organization’s current practices, to make sure we deliver a solution that fits your business model, culture, and operations, as well as improving the organization’s decisions. For example, a decision-support solution we developed for a major shipping company has saved the customer about five million dollars per year in operational expenses, for each of the past eight years, delivering an estimated twenty-fold return on investment.
See “What makes Mosaic better” above. We’re not the right consultants for every problem. We are unsurpassed in our ability to solve complex business optimization problems. Our customers and partners still work with us after a decade of experience because we consistently deliver superlative results. If you have a complex business problem involving substantial data, we can do the same for you.